Fs

Environ. Sci. Technol. 2010, 44, 5409–5415

Polychlorinated Dibenzo-p-Dioxins and Dibenzofurans (PCDD/Fs) Impurities in Pesticides: A Neglected Source of Contemporary Relevance E V A H O L T , * ,† R O L A N D W E B E R , ‡ GAVIN STEVENSON,§ AND C A R O L I N E G A U S * ,† The University of Queensland, National Research Centre for Environmental Toxicology (EnTox), 39 Kessels Road Coopers Plains, QLD 4108, Australia, POPs Environmental Consulting, 73035 Go¨ppingen, Germany, and The National Measurement Institute, Dioxin Analysis Unit, Pymble, NSW 2072, Australia

Received December 24, 2009. Revised manuscript received June 7, 2010. Accepted June 7, 2010.

Polychlorinated dibenzo-p-dioxins and dibenzofurans (PCDD/ Fs) may be formed during the manufacture of chlorinated pesticides, and can remain in the products as impurities. However, the contemporary release of PCDD/Fs to the environment from pesticide use is poorly understood. For this study, 27 pesticide formulations were analyzed for PCDD/Fs (n ) 23 registered for use in Australia). PCDD/F impurities were present in all samples, ranging from 0.020 to 2100 ng ΣPCDD/F g-1 active ingredient (AI). Among current use pesticides, pentachloronitrobenzene (PCNB) contained the highest impurity levels (up to 2000 ng ΣPCDD/F g-1 AI and 5.6 ng TEQ g-1 AI). The quantity of pesticide used in Australia and associated release of PCDD/Fs was estimated for PCNB and phenoxy herbicides (2,4-dichlorophenoxyacetic acid (2,4-D) and 2,4-dichlorophenoxybutyric acid (2,4-DB)) using a probabilistic approach. Input parameters to model pesticide use contributed the highest proportions to the variability of the estimated PCDD/F release, and were considered to have the highest uncertainty. Preliminary estimates of PCDD/F release suggest that contaminated pesticides represent an important ongoing PCDD/F source to the Australian environment (10th-90th percentiles for PCNB ) 14-42 and 2,4-D/2,4-DB ) 0.35-1.6 g TEQ annum-1). These results may have global relevance given that many of the pesticides analyzed were imported into Australia, and are used in high volumes in other countries.

Introduction The production and disposal of chlorinated chemicals, including pesticides, and their waste streams represent wellknown sources of polychlorinated dibenzo-p-dioxins and dibenzofurans (PCDD/F) to the environment (1, 2). In addition, the unintentional formation of PCDD/Fs during pesticide production can lead to contamination of the pro* Address correspondence to either author. Telephone: +61 7 3274 9004; fax: +61 7 3274 9003; e-mail: [email protected]; c.gaus@ uq.edu.au. † The University of Queensland. ‡ POPs Environmental Consulting. § The National Measurement Institute. 10.1021/es903915k

 2010 American Chemical Society

Published on Web 06/18/2010

duct, thus providing a pathway for PCDD/F release to the environment, and widespread distribution during pesticide application (3). Pesticides that have been repeatedly shown to contain high PCDD/F impurities (>1 000 up to 15 000 000 ng ΣPCDD/F g-1) include pentachlorophenol (PCP) (3-5), chloronitrofen (CNP) (3) and 2,4,5-trichlorophenoxy acetic acid (2,4,5-T) (6-8). Although these pesticides have been phased out or are banned in many countries today, the historical use of these products has resulted in considerable environmental contamination with PCDD/Fs. For example, the use of PCP in Japanese rice paddies during 1955-1974 released an estimated 80 000-1 000 000 kg ΣPCDD/Fs or 220-270 kg TEQWHO98 (3). The use of contaminated 2,4,5-T based formulations (e.g., Agent Orange) during the Vietnam War represents another well-known example for the release of high amounts of PCDD/Fs (9). It has been estimated that 366 kg of 2,3,7,8-tetrachlorodibenzo-p-dioxin (2,3,7,8-TCDD) was released in Vietnam via the spraying of several formulations containing 2,4,5-T (9). Due to the high persistency of these contaminants in the environment, such historical PCDD/F releases can represent secondary sources to the contemporary environment via long-term volatilization, particle-bound transport, and uptake into the food chain. In Sweden, PCDD/F contamination from the past use of chlorophenols at wood treatment facilities (400-500 sites with a total estimated load in soil of 5-10 kg TEQ) still represents a significant secondary PCDD/F source, even after a 30-year ban on the use of these fungicides (10, 11). In addition to such well-known historical cases (which are mainly focused on 2,4,5-T and PCP), other pesticides reported to contain PCDD/Fs include some (i) chlorophenols (e.g., tetrachlorophenol 4, 11) and (ii) phenoxy herbicides (e.g., 2,4-dichlorophenoxyacetic acid (2,4,-D) 3, 4, 12, 13), as well as a few representatives of (iii) chlorinated herbicides (e.g., chloronitrofen 3, 14), (iv), chlorinated fungicides (e.g., chlorothalonil (3)), and (v) organochlorines (e.g., lindane (γHCH) (4)). Analytical data for 2,4-D (0.0039-273 ng ΣPCDD/ Fs g-1 3, 4, 12, 13) and chlorothalonil (3.6-26 ng ΣPCDD/Fs g-1 (3)) suggest that PCDD/F concentrations in such formulations are generally lower than in pesticides that have been largely banned or severely restricted in the 1980/90s. Additionally, improvements of production technologies and practices since the 1990s are thought to have resulted in a decrease of PCDD/F impurities (4). It is expected that this in turn would affect a decrease in the associated release from the use of pesticides, assuming the quantity of PCDD/F contaminated pesticides used has not increased significantly. However, there is a distinct lack of publicly available information on contemporary PCDD/F levels in pesticides, even for those used at high volume (e.g., 2,4-D), and thus the associated contemporary PCDD/F release cannot be quantified. Of the little publicly available information on PCDD/F impurity levels and release with pesticides, the USEPA provides among the most comprehensive data (4). In their most recent evaluation on chemical PCDD/F sources, USEPA judged the pesticide 2,4-D as the only product with the potential for environmental release through agricultural use (4). It has also been reported that most manufacturers of 2,4-D formulations in the United States have taken steps to significantly reduce the level of PCDD/F impurities since the mid 1990s. However, PCDD/F release from 2,4-D use in the U.S. could not be estimated since 1995 due to the lack of more recent analytical data (4). In addition, high volume use of formulations contaminated even with relatively low levels of PCDD/Fs may still result in significant release of PCDD/ Fs into the environment. For example, PCDD/F releases from VOL. 44, NO. 14, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

5409

the widespread use of 2,4-D in the U.S. still represented the second most important source of dioxins to land in 1995 (18 g I-TEQ) (4, 15, 16). The USEPA lists 161 pesticides that have the potential to contain PCDD/F impurities when manufactured under conditions that favor dioxin formation (4). For the majority of these, the manufacturing processes have been evaluated for potential PCDD/F formation but analytical data are not (publicly) available (4). Levels of PCDD/F impurities can, however, vary markedly among pesticide products indicating that measures to control PCDD/F formation during pesticide production may differ among manufacturing plants, years, batches, and active ingredients (3, 5). In addition, some of the pesticides with high potential to contain PCDD/F impurities (e.g., PCP) are not banned in all countries, although their use is often restricted (17-19). Due to the lack of information regarding PCDD/F impurity levels in currently used pesticides, their potential to represent contemporary PCDD/F sources cannot be evaluated. This is reflected in PCDD/F emissions inventories, which rarely include contributions of current-use pesticides to overall releases of PCDD/F, including in Australia. Sixty percent of Australia’s land is used for agriculture (20), coinciding with a long history of extensive pesticide use (21), in particular as part of intensive agricultural land-use within the coastal plains. Australia has no regulatory limit for PCDD/F levels in current use pesticides and no regulatory monitoring and reporting is undertaken. Among the few historical data available on PCDD/F impurities in Australian pesticides, data quality is poor with no reported or high limits of quantification (LOQ). For example, PCDD/F impurities in 2,4-D have been reported below the LOQ in samples analyzed prior to 1993, however the LOQ value was not provided (22). Similarly, PCDD/F impurities in mecoprop are reported below LOQ, with the LOQ ranging from 0.1 to 1000 ppb (23). In contrast, a review of the production process of dichlorprop-P reportedly did not warrant analytical monitoring for PCDD/F contamination (24). 2,4,5-T, which is today banned from use in Australia, was the only pesticide for which PCDD/F levels were monitored and regulated. Analysis of 2,4,5-T containing formulations (n ) 121) between 1975 and 1985 determined that 6 formulations contained PCDD/F levels by up to a factor of 8 above the then regulatory limit of 100 ng TCDD g-1 (reported in ref 25), however, no analyzed samples (n ) 19) exceeded the revised limit after 1982 (10 ng TCDD g-1) (25). This study was undertaken to investigate the presence of PCDD/Fs in current use pesticides and to improve available information with respect to PCDD/F impurity levels in both current and past use pesticides. Different herbicide, insecticide, and fungicide formulations were analyzed for PCDD/ Fs, including a range of active ingredients which have never been tested and reported previously. The results from these analyses were used to estimate PCDD/F release associated with selected pesticide formulations, and to evaluate their potential significance compared to other known PCDD/F emission sources in Australia.

Materials and Methods Sample Details. Twenty-seven pesticide formulations (with 18 different primary active ingredients) were selected for PCDD/F analysis (Figure S1, Supporting Information (SI)). Pesticide classes and active ingredients included different phenoxy herbicides (2-methyl-4-chlorophenoxy-acetic acid (MCPA), 2,4-dichlorophenoxyacetic acid (2,4-D), 2,4-dichlorophenoxybutyric acid (2,4-DB), and 2,4,5-trichlorophenoxyacetic acid (2,4,5-T)), fungicides (prochloraz, chlorothalonil, and PCNB, also known as quintozene), insecticides (chlorpyrifos, fenamiphos, lindane, heptachlor, and chlordane), and other herbicides (chlorthal, fluoroxypyr, triclopyr, 5410

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 14, 2010

mecoprop, imazamox, and flumetsulam). These represent mainly agrochemical formulations (n ) 22) but also some household products (n ) 5), and include those with active ingredients registered for use in Australia at the time of this study (n ) 23), as well as some that are obsolete (i.e., no longer registered, banned from use, or no longer manufactured today in most countries, including Australia; n ) 4). Each pesticide formulation is referred to by its active ingredient. Where different formulations with the same active ingredient were analyzed, samples are denoted with letters (e.g., PCNB(a); PCNB(b), and PCNB(c) represent three different products). Replicate analysis (i.e., the same formulation was analyzed twice) is denoted by “Replicate” (e.g., 2,4-D(a) Replicate). Information regarding the products (e.g., date of manufacture, matrix, use status in Australia, active ingredient and its concentration), where available, is listed in Table S1, SI. Formulations were obtained from a variety of sources, including a herbicide mixing/manufacturing plant (n ) 2), farmer (n ) 1), government department (n ) 8), horticulture/ farm suppliers (n ) 9), and a hazardous waste disposal facility (n ) 7). To facilitate comparisons to technical products analyzed in other studies and between different formulations, PCDD/F concentrations were normalized to active ingredient (AI), in addition to reporting concentrations on whole formulation basis (FL). Analysis for PCDD/Fs. Two different extraction techniques were employed depending on the pesticide matrix (liquid or solid). A summary of the analytical methodology is outlined in Figure S2, SI. For pesticides in liquid form, PCDD/Fs were extracted using a method routinely employed for industrial oil samples (26), previously adapted from ref 27. In brief, a known quantity (∼2 g) of liquid pesticide formulations (except for 2,4,5-T/2,4-D mix; sample size 100 µL) was dissolved in approximately 20 mL of n-hexane followed by liquid/liquid partitioning into dimethyl sulphoxide (DMSO). The DMSO fraction was diluted with 500 mL of Milli-Q water and then re-extracted into n-hexane. For solid pesticides, a known quantity of sample (∼2 g) (except chlordane and replicate analyses of PCNB pesticide formulation; sample sizes 0.1 and 0.06 g, respectively) was subject to automated Soxhlet extraction with toluene for 8 h and solvent exchanged to n-hexane, according to methods described in refs 26 and 28. Crude extracts of all samples underwent the same cleanup process, commencing with an acid preclean using pure concentrated sulphuric acid. Samples were then purified using Powerprep automated system (Fluid Management Systems, Waltman, MA) (26). In brief, sample extracts were applied to multilayered (acid and base modified) silica-basic alumina column in sequence and eluted with hexane, followed by dichloromethane/hexane (2:98) and dichloromethane/hexane (50:50). The 50:50 dichloromethane/ hexane solvent mix was transferred to a carbon column which was then eluted by ethyl acetate/toluene (50:50) in the forward direction and toluene in the reverse direction. The toluene fraction was collected for all samples and concentrated for high-resolution gas chromatography/high-resolution mass spectrometry (HRGC/HRMS). HRGC/HRMS was conducted on a MAT95XL HRMS (ThermoFinnigan MAT GmbH, Bremen, Germany) and an Agilent GC (Palo Alto, CA) equipped with a CTC autosampler. A ZB-5 MS (Phenomenex, Lane Cove, Australia) capillary column was operated under temperature programmed conditions of 100 °C for 1 min, 100-200 at 40 °C min-1 followed by 200-235 at 3 °C min-1 and 235-310 at 5 °C min-1. The injection port temperature was maintained at 290 °C. GC carrier gas was helium maintained at a flow rate of 1.0 mL/min. The transfer line was maintained at 280 °C. Mass spectra were recorded with an electron impact ionization source operated at 240 °C and 70 eV and the mass

spectrometer was operated on a resolution of approximately 10 000 (10% valley). PCDD/Fs were quantified according to isotope dilution techniques using a surrogate standard that included the 15 13C12-labeled 2,3,7,8-subsituted PCDD/Fs (Wellington Laboratories, Ontario, Canada) listed in the SI (Table S2); for quantification of OCDF and 1,2,3,7,8,9-HxCDD, the 13C12 OCDD and average response of the labeled analogues of the other two 2,3,7,8-HxCDD surrogate standards, respectively were used, as described in refs 26 and 28. Two 13 C12-labeled PCDD/Fs (Wellington Laboratories, Ontario, Canada) were used as recovery standards (Table S2, SI) and PCDD/F quantities in the sample were corrected for their respective surrogate recoveries. Toxic equivalencies (TEQs) were calculated using the most recent toxic equivalency factors (TEFs) adopted by WHO (29) and reported using lower bound values unless stated otherwise; where literature data could not be converted, TEQ levels based on other published schemes (e.g., I-TEQ) are provided and identified as such. Quality Assurance/Quality Control. A detailed description of instrument calibration, quality control criteria for identification and quantification of PCDD/Fs, and quality assurance (e.g., blanks, limit of detection (LOD), recoveries, monitoring for interferences) is provided in the SI, Section 1; and Tables S2 and S3. In brief, the majority of PCDD/Fs were below the LOQ in batch blanks, however, some congeners were present (in most cases OCDD) at low concentrations (0.00010-0.063 ng g-1). Where quantification criteria were not met by individual congeners, these are annotated in Table S1 (SI) and were treated as below the limit of quantification. Seven samples had relatively low recoveries for some or all PCDD/F congeners (7-25%) (Section 1, S1). These are listed in comparison to USEPA labeled compound recovery limits (29) in the SI (Table S3) and have been annotated in Table S1, SI. The calculated concentration of these congeners must be considered to have higher uncertainty; recovery criteria were met by all samples used for estimating PCDD/F release. Information on the variability of analytical procedures (replicate analysis) is also provided in the SI, Table S1 and Section S1. Estimating PCDD/F Release from Pesticide Use. To quantify the environmental release of PCDD/Fs as pesticide impurities, the amount of pesticides used needs to be taken into account. Data on the use, import, or sales of individual pesticide active ingredients or formulations are, however, commercial in confidence in Australia and thus not available to the public (21). For the present study, a probabilistic approach was used to model the annual use of PCNB, and 2,4-D/2,4-DB (phenoxy herbicides with a similar structure, mode of action and usage) in Australia, using (i) pesticide label application rates, (ii) application frequencies, (iii) Australian crop and land-use areas, and (iv) the percent crop treated (Tables S4 and S5, SI). These active ingredients where selected as PCNB contained the highest levels of PCDD/Fs among current use pesticides, and both PCNB and 2,4-D (and its various forms) represent the most commonly used fungicides and herbicides (respectively) in Australia (21). Crystal Ball risk modeling software (Oracle; Version 11.1.1.3.00) was used to assign probability distributions to each parameter (Table S6, SI). Where no information on the most likely value was available (e.g., application rates), reported minima and maxima were used to define the range for uniform distributions. Where no minimum or maximum values were publically available (e.g., frequency of application, percent area treated), a relatively wide, likely range was estimated and uniform distributions were adopted. For parameters where the mean and its variability could be obtained (e.g., crop area) normal distributions were adopted. Because of the relatively small sample sizes for PCDD/F analyses, conventional distribution fitting techniques (such as goodness of fit tests and bootstrapping) for TEQ impurity

levels in pesticides analyzed for the present study could not be employed. A normal distribution (truncated at zero) was adopted for middle bound TEQ concentrations, using the calculated mean and standard deviation multiplied by two (to account for higher variability) in pesticide formulations containing PCNB (n ) 3) and 2,4-D/2,4-DB (n ) 4). Monte Carlo simulation was then employed to generate a probability forecast on the annual quantity of pesticide use, and the associated TEQ release in Australia, by repeatedly (10 000 iterations) and randomly drawing values from each input distribution. Sensitivity analysis was conducted to identify the most sensitive input parameters for consideration of their respective uncertainty and influence on overall release estimates. For 2,4,5-T, only one formulation was analyzed for the present study. The historical PCDD/F release associated with the past use of this pesticide in Australia was estimated by using previously published information (30) (approximately 300 tonnes annum-1 active ingredient between 1969 and 1971) and the middle bound TEQ from the present study.

Results and Discussion Levels of PCDD/Fs in Pesticide Formulations. PCDD/Fs were detected in all 23 current use as well as the 4 obsolete pesticide formulations analyzed for this study (Figure 1 and Table S1, SI) which encompass a variety of insecticides, herbicides, and fungicides. This indicates that the presence of PCDD/F impurities is common in a wide range of pesticides. PCDD/F concentrations in these formulations ranged from relatively low (0.0043 and 0.020 ng ΣPCDD/F g-1 based on formulation (FL) and active ingredient (AI), respectively) to high levels (1500 and 2100 ΣPCDD/F ng-1 based on FL and AI, respectively). Highest PCDD/F impurity levels were present in obsolete formulations containing 2,4,5T/2,4-D (2100 ng ΣPCDD/F g-1 AI) but also in current use pesticides containing the active ingredient PCNB (also known as quintozene: average 1500, range 1100-2000 ng g-1 AI, n ) 3 formulations), followed by 2,4-DB, chlorothalonil, lindane, chlorthal, 2,4-D, chlordane/heptachlor, and chlordane in order of decreasing concentrations (Figure 1). TEQ levels in the three PCNB formulations were also elevated with average lower bound concentrations of 3.9, ranging from 2.4 to 5.6 ng g-1 AI, respectively (Figure 1 and Table S1, SI). Similarly, TEQ levels were elevated in the obsolete formulation with 2,4,5-T/2,4-D as active ingredient (85 ng g-1 AI). The remaining pesticides contained TEQ levels below 0.43 ng g-1 AI. Although some of these pesticides are no longer used in many countries, 15 of the 18 active ingredients analyzed in this study (e.g., PCNB, 2,4-DB, chlorothalonil, chlorthal, 2,4-D; Figure 1) are currently registered for use in Australia (31) and elsewhere (e.g., U.S. (17), Africa (18), and Europe 32, 33). This demonstrates that elevated PCDD/F levels can still be present in current use formulations and are not restricted to the few obsolete pesticides, which are well-known for their potential to contain high PCDD/F impurities, such as 2,4,5-T (6, 34). The commonly used fungicide PCNB (17-19, 21, 35), which contained the highest PCDD/F impurity levels among the current use pesticides analyzed for this study, is not included in the USEPA list of pesticides with the potential to contain dioxin impurities (4). To the knowledge of the authors, no other data on PCDD/F impurities have been published for PCNB, however, a previous report indicates monitoring had been carried out in the 1990s by the Environment Agency of Japan (cited in ref 3). The cited report suggests the Japanese formulations contained approximately 20-30 fold lower TEQ levels compared to those of the present study. Analytical data for PCDD/F impurities in fluoroxypyr, flumetsulam, triclopyr, chlorthal, imazamox, mecoprop, fenamiphos, chlorpyrifos, and prochloraz have not been VOL. 44, NO. 14, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

5411

FIGURE 1. PCDD/F concentrations (ng g-1) and lower bound TEQ (ng g-1) in pesticides based on formulation and active ingredient (percent active ingredient was not available for chlordane and chlordane/heptachlor). Average values are presented where replicate analysis was conducted (n ) 2); Replicate data available in Table S1, SI. Obsolete (banned or redundant) active constituents are marked (2). All other pesticides are currently used in Australia (the date of manufacture is provided in Table S1, SI). reported previously. In the present study, PCDD/F levels in these pesticides (ranging from 0.054 to 62 ng ΣPCDD/F g-1AI) were comparable to those detected in pesticides known for PCDD/F contamination (MCPA: 0.020-3.9 ng ΣPCDD/F g-1AI and 2,4-D: 0.51-10 ng ΣPCDD/F g-1AI; this study). TEQ levels in these pesticides were relatively low (not detected (nd) to 0.059 ng TEQ g-1AI), however, some of these products are among the most commonly used pesticides in Australia (21). Thus, these pesticides represent previously unknown, ongoing sources of PCDD/Fs to the environment which have never been evaluated. Similarly, very limited analytical data is published for the remaining current use pesticides. The few data available for chlorothalonil (7.2 and 35 ng g-1 AI (3)) or lindane (2 ng g-1 AI (12)), are comparable to results obtained for this study (chlorothalonil: 39 ng ΣPCDD/F g-1AI and lindane: 1.4 - 19 ng ΣPCDD/F g-1AI), but have never been evaluated as sources for PCDD/Fs to the environment. Pesticides chlordane and heptachlor, now banned from use in many countries, have also never been reported to contain PCDD/Fs, but were found to contain relatively high impurity levels up to 24 ng ΣPCDD/F g-1 FL in the present study (no information was available on the percent active ingredient). Given that both heptachlor and chlordane were used in high volumes (e.g., >1000 tonnes of chlordane was used in the U.S. (36); no data available for Australia) and that PCDD/F are highly persistent, the past use of these pesticides may represent a secondary (i.e., reservoir) source of PCDD/Fs to the contemporary environment that has not yet been evaluated. For pesticides that are well-known for their potential to contain PCDD/F impurities, for example 2,4-D, improvements of production technologies and practices are generally 5412

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 14, 2010

expected to minimize or prevent contamination of contemporary products (16, 22). A study from Japan has, for example, shown that PCDD/F levels in pesticides with the same active ingredient (2,4-D, PCP, chloronitrofen, and chlorothalonil) generally decreased chronologically between years of production (3). In the present study, however, 2,4-D formulations (2,4-D(a) and 2,4-D(b)) manufactured in 2006 and 2005, respectively), contained TEQ levels (0.00098-0.17 ng TEQ g-1 AI) comparable to those manufactured 10-20 years ago (0.0019-0.83 ng I-TEQ g-1 3, 4, 13). These results indicate that reduction measures to avoid PCDD/F impurities in pesticides are not applied or effective at all pesticide manufacturing plants. The obsolete pesticides containing 2,4,5-T and mixtures of 2,4,5-T/2,4-D (e.g., Agent Orange) have been shown repeatedly to contain PCDD/F impurities, particularly 2,3,7,8-TCDD. The PCDD/F concentration in 2,4,5-T/2,4-D (340 ng ΣPCDD/F g-1 FL and 2100 ng ΣPCDD/F g-1 AI) analyzed as part of this study was within the range of those reported previously (100-5100 ng ΣPCDD/F g-1 6, 12), and contained similar levels of 2,3,7,8-TCDD (81 ng 2,3,7,8-TCDD g-1 AI) compared to a 1977 product previously analyzed from the same manufacturer in Australia (150 ng 2,3,7,8-TCDD g-1 AI) (25). Estimated PCDD/F Release from Pesticides Used in Australia. Probability distributions on the current annual use of PCNB and 2,4-D/2,4-DB in Australia, and their impurity levels, were generated to provide a first estimate on the PCDD/F release associated with the use of these pesticides (see Materials and Methods and Tables S4 and S5, SI). The estimated average annual TEQ release to the Australian environment was 27 g (10th-90th percentiles: 14-42) for

FIGURE 2. Estimated PCDD/F releases from current-use pesticides PCNB and 2,4-D/2,4-DB and historical use of 2,4,5-T (left) compared to the top 19 PCDD/F sources to land from the Australian dioxin inventory (40) and re-evaluations for biomass burning* (41). Values for sources listed under the Australian inventory represent the best estimate. [†Estimated PCDD/F release associated with 2,4,5-T/2,4-D does not consider historical variations in impurity levels over its long-term use (see discussion). *Estimated release from biomass burning is based on the recently re-evaluated estimates (41); previous estimates from the Australian inventory (40) were 1020 g annum-1. All estimates for the Australian emission inventory were based on TEFs from 1998.] PCNB and 0.87 g (10th-90th percentiles: 0.35-1.6) for 2,4D/2,4-DB (Figure 2 and Tables S4 and S5, SI). Sensitivity analysis indicated that overall, input parameters to quantify pesticide use contributed the highest proportions to the variability of the estimated PCDD/F release (Table S6, SI). These parameters (specifically the percent area treated and frequency of application) were also considered to have the highest uncertainty (Table S6, SI). A review on rural pesticide use in Australia from 1996 to 1999 indicated that PCNB, which was listed among the most important individual fungicides, may have been used at >10, >100, or >1000 tonnes AI annum-1 for agricultural applications (21). Similarly, rural agricultural use of >1000 tonnes AI annum-1 of plant cell growth disrupting chemicals was estimated for these years, including the herbicide 2,4-D as among the major chemicals in this group (21). For the present study, average agricultural use of PCNB and 2,4-D was forecast at approximately 2200 (10th-90th percentiles: 1300-3200) and 4400 (10th-90th percentiles: 2200-7100) tonnes AI annum-1 (average), respectively (Tables S4 and S5, SI). Whereas total product sales for both fungicides and herbicides in Australia have increased over the past 10 years (37, 38), current (or more precise historical) estimates for agricultural (or nonagricultural) use of PCNB and 2,4-D/2,4-DB could not be obtained to quantify the uncertainty of the present results. For PCNB, nonagricultural applications contributed the highest proportion (65%) to its estimated use and the associated PCDD/F release (Table S4, SI). Given the high sensitivity and uncertainty of several input parameters to model pesticide use in Australia, the average PCDD/F release estimated for the present study must be regarded as uncertain and should be considered in context with the provided ranges (Figure 2 and Tables S4 and S5, SI). Quantitative information

on pesticide active ingredients used in Australia is required to further evaluate the PCDD/F release associated with pesticides, and would facilitate decision making on management strategies for source elimination. In addition to uncertainties regarding pesticide use, the PCDD/F release estimate of the present study assumes that the impurity data for the three and four formulations containing PCNB and 2,4-D/2,4-DB (respectively) are representative of other formulations currently used in Australia (i.e., it was assumed that the TEQ of approximately 68% of marketed formulations is within 2 standard deviations of the calculated mean). Although no other data are currently available to assess the variability of TEQ levels across more formulations, the Australian federal government has been alerted to the present findings and has, after conducting their own analysis, recently suspended further registration of PCNB on the basis of PCDD/F contamination (39). This suggests that elevated TEQ levels in PCNB formulations have been independently confirmed. Prior to this study, there has been no consideration of pesticide use as a source of PCDD/Fs to the Australian environment. Consequently, these sources have not been subject to measures for emission reduction. Despite the uncertainties associated with the present PCDD/F release estimates, even consideration of the lower ranges (i.e., 10th percentiles and minima) suggests that pesticides represent an important ongoing source of PCDD/Fs. Under a worst case scenario (considering the 90th percentile or maximum release estimates), the PCDD/F release from the use of PCNB alone may be near or higher compared to identified priority sources in Australia (Figure 2) as reported in the most recent Australian PCDD/F emissions inventory (40) and subsequent re-evaluations (41). The combined PCDD/F release of these VOL. 44, NO. 14, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

5413

and other pesticides shown here, or with the potential to contain PCDD/F impurities, should thus be further evaluated and quantified. PCDD/F release estimated for the obsolete pesticide 2,4,5-T ranged from 13 to 27 g TEQ annum-1 based on reported annual use in Australia between 1969 and 1971 (∼300 tonnes annum-1) as reported by ref 30 and PCDD/F levels obtained in this study. However, TEQ impurity levels are likely to have fluctuated over the period of registration in Australia (approximately from the 1960s to the late 1980s 25, 42). For example, during compliance testing in Western Australia (1975-1982) PCDD/F levels in 2,4,5-T varied from